DATA_PATH <- here("data/processed/syntactic_bootstrapping_tidy_data.csv") # make all variables (i.e. things that might change) as capital letters at the top of the scripts

ma_data <- read_csv(DATA_PATH) 
ma_data <- ma_data %>% filter(paradigm_type == "action_matching")

Data Overview

action

n_effect_sizes <- ma_data %>%
  filter(!is.na(d_calc)) %>%
  nrow()

n_papers <- ma_data %>%
  distinct(unique_id) %>%
  nrow()

There are 79 effect sizes collected from 22 different papers.

Here are the papers in this analysis:

ma_data %>%
  count(short_cite) %>%
  arrange(-n) %>%
  DT::datatable()

# Forest plot

ma_model <- rma(ma_data$d_calc, ma_data$d_var_calc)
ma_model
## 
## Random-Effects Model (k = 79; tau^2 estimator: REML)
## 
## tau^2 (estimated amount of total heterogeneity): 0.6446 (SE = 0.1243)
## tau (square root of estimated tau^2 value):      0.8029
## I^2 (total heterogeneity / total variability):   88.11%
## H^2 (total variability / sampling variability):  8.41
## 
## Test for Heterogeneity:
## Q(df = 78) = 409.5470, p-val < .0001
## 
## Model Results:
## 
## estimate      se    zval    pval   ci.lb   ci.ub 
##   0.5414  0.0999  5.4175  <.0001  0.3456  0.7373  *** 
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
forest(ma_model,
       header = T,
       slab = ma_data$unique_id,
       col = "red",
       cex = .7)

#funnel plot {.tabset} ## vanilla

ma_data %>% 
  mutate(color = ifelse(sentence_structure == "transitive", "red", "blue"),
         color_sample_size = ifelse(n_1 < 10, "red", ifelse(n_1 < 20, "orange", "yellow")),
         color_confidence = ifelse(inclusion_certainty == 1, "red", "black"))-> ma_data_funnel


ma_data_funnel %>% filter (abs(d_calc) < 5) -> ma_data_funnel_no_outlier

ss_colors <- ma_data_funnel$color
ss_colors_no_outlier <- ma_data_funnel_no_outlier$color 

ma_model_funnel <- rma(ma_data_funnel$d_calc, ma_data_funnel$d_var_calc)
ma_model_funnel_no_outlier <- rma(ma_data_funnel_no_outlier$d_calc, ma_data_funnel_no_outlier$d_var_calc)

f1<- funnel(ma_model_funnel, xlab = "Effect Size", col = ss_colors) 
legend("topright",bg = "white",legend = c("transitive","intransitive"),pch=16,col=c("red", "blue"))
title(main = "All effect sizes break down by sentence structure")

f2<- funnel(ma_model_funnel_no_outlier, xlab = "Effect Size", col = ss_colors_no_outlier) 
legend("topright",bg = "white",legend = c("transitive","intransitive"),pch=16,col=c("red", "blue"))
title(main = "effect sizes excluded outliers (abs <5) break down by sentence structure")

take int occount sentence structure?

ma_model_sentence_structure <- rma(ma_data_funnel$d_calc~ma_data_funnel$sentence_structure, ma_data_funnel$d_var_calc)
ma_model_no_outlier_ss <- rma(ma_data_funnel_no_outlier$d_calc~ma_data_funnel_no_outlier$sentence_structure, ma_data_funnel_no_outlier$d_var_calc)

f3 <- funnel(ma_model_sentence_structure, xlab = "effect size", col = ss_colors)

f3_b <- funnel(ma_model_no_outlier_ss, xlab = "effect size", col = ss_colors_no_outlier)

take int account sentence structure and age?

ma_model_ss_age <- rma(ma_data_funnel$d_calc~ma_data_funnel$sentence_structure + ma_data_funnel $mean_age, ma_data_funnel$d_var_calc)
ma_model_funnel_no_outlier_ss_age <- rma(ma_data_funnel_no_outlier$d_calc~ma_data_funnel_no_outlier$sentence_structure+ma_data_funnel_no_outlier$mean_age, ma_data_funnel_no_outlier$d_var_calc)



f4 <- funnel(ma_model_ss_age, xlab = "effect size", col = ss_colors)

f4_b <- funnel(ma_model_funnel_no_outlier_ss_age, xlab = "effect size", col = ss_colors_no_outlier)

Variable Summary

Continuous Variables

CONTINUOUS_VARS <- c("n_1", "x_1", "sd_1", "d_calc", "d_var_calc", "mean_age")

long_continuous <- ma_data %>%
  pivot_longer(cols = CONTINUOUS_VARS)

long_continuous %>%
  ggplot(aes(x = value)) +
  geom_histogram() + 
  facet_wrap(~ name, scale = "free_x") +
  labs(title = "Distribution of continuous measures")

long_continuous %>%
  group_by(name) %>%
  summarize(mean = mean(value),
            sd = sd(value)) %>%
  kable()
name mean sd
d_calc 0.5577306 1.4076258
d_var_calc 0.2043085 0.3694061
mean_age 906.7969722 338.3888715
n_1 14.8734177 6.1880097
sd_1 0.1417692 0.0751996
x_1 0.5659904 0.0950815

Categorical Variables

CATEGORICAL_VARS <- c("sentence_structure", "agent_argument_type", "patient_argument_type", "stimuli_actor",
                     "presentation_type", "character_identification",
                     "test_mass_or_distributed", "practice_phase", "test_method")

long_categorical <- ma_data %>%
  pivot_longer(cols = CATEGORICAL_VARS) %>%
  count(name, value) # this is a short cut for group_by() %>% summarize(count = n()) 

long_categorical %>%
  ggplot(aes(x = value, y = n)) +  
  facet_wrap(~ name, scale = "free_x") +
  geom_col(position = 'dodge',width=0.4) + 
  theme(text = element_text(size=8),
        axis.text.x = element_text(angle = 90, hjust = 1))  # rotate x-axis text

Explore Moderators

no moderators

all

m1 <- rma.mv(d_calc, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data)
summary(m1)
## 
## Multivariate Meta-Analysis Model (k = 79; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -163.3487   326.6973   330.6973   335.4107   330.8573   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.2147  0.4634     22     no  short_cite 
## 
## Test for Heterogeneity:
## Q(df = 78) = 409.5470, p-val < .0001
## 
## Model Results:
## 
## estimate      se    zval    pval   ci.lb   ci.ub 
##   0.3843  0.1080  3.5574  0.0004  0.1726  0.5961  *** 
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

<36

ma_data_young <- ma_data %>%
    mutate(age_months = mean_age/30.44) %>% 
    filter(age_months < 36) 

m_young <- rma.mv(d_calc, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_young)
summary(m_young)
## 
## Multivariate Meta-Analysis Model (k = 57; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -115.7270   231.4539   235.4539   239.5046   235.6803   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.2940  0.5422     17     no  short_cite 
## 
## Test for Heterogeneity:
## Q(df = 56) = 302.2290, p-val < .0001
## 
## Model Results:
## 
## estimate      se    zval    pval   ci.lb   ci.ub 
##   0.4071  0.1414  2.8791  0.0040  0.1299  0.6842  ** 
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

>= 36

ma_data_old <- ma_data %>%
    mutate(age_months = mean_age/30.44) %>% 
    filter(age_months > 36 | age_months == 36) 

m_old <- rma.mv(d_calc, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_old)
summary(m_old)
## 
## Multivariate Meta-Analysis Model (k = 22; method: REML)
## 
##   logLik  Deviance       AIC       BIC      AICc 
## -42.9215   85.8431   89.8431   91.9321   90.5098   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.0278  0.1668      6     no  short_cite 
## 
## Test for Heterogeneity:
## Q(df = 21) = 107.2915, p-val < .0001
## 
## Model Results:
## 
## estimate      se    zval    pval   ci.lb   ci.ub 
##   0.3074  0.0947  3.2448  0.0012  0.1217  0.4931  ** 
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

age only

colored by unique_id

ma_data %>% 
  ggplot(aes(x = mean_age/30.44, y = d_calc,color = unique_id)) +
  geom_point() +
  ylab("Effect Size") +
  xlab("Age (days)") +
  ggtitle("Syntactical Bootstrapping effect size vs. Age (months)") 

all

ma_data %>% 
  ggplot(aes(x = mean_age/30.44, y = d_calc, size = n_1)) +
  geom_point() +
  geom_smooth(method = "lm") +
  geom_smooth(color = "red") +
  ylab("Effect Size") +
  xlab("Age (days)") +
  ggtitle("Syntactical Bootstrapping effect size vs. Age (months)") +
  theme(legend.position = "none") 

m_simple <- rma.mv(d_calc ~ 1, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data)

m_age <- rma.mv(d_calc ~ mean_age, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data)
summary(m_simple)
## 
## Multivariate Meta-Analysis Model (k = 79; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -163.3487   326.6973   330.6973   335.4107   330.8573   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.2147  0.4634     22     no  short_cite 
## 
## Test for Heterogeneity:
## Q(df = 78) = 409.5470, p-val < .0001
## 
## Model Results:
## 
## estimate      se    zval    pval   ci.lb   ci.ub 
##   0.3843  0.1080  3.5574  0.0004  0.1726  0.5961  *** 
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
summary(m_age)
## 
## Multivariate Meta-Analysis Model (k = 79; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -159.9705   319.9409   325.9409   332.9723   326.2697   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.2182  0.4671     22     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 77) = 408.4665, p-val < .0001
## 
## Test of Moderators (coefficient 2):
## QM(df = 1) = 6.2052, p-val = 0.0127
## 
## Model Results:
## 
##           estimate      se     zval    pval    ci.lb    ci.ub 
## intrcpt     0.7840  0.1938   4.0458  <.0001   0.4042   1.1638  *** 
## mean_age   -0.0004  0.0002  -2.4910  0.0127  -0.0008  -0.0001    * 
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
forest(m_age,
       header = T,
       slab = ma_data$unique_id,
       col = "red",
       cex = .7
)

funnel(m_age)

young, < 36

ma_data_young %>%
  ggplot(aes(x = age_months, y = d_calc, size = n_1)) +
  geom_point() +
  geom_smooth(method = "lm") +
  geom_smooth(color = "red") +
  ylab("Effect Size") +
  xlab("Age (days)") +
  ggtitle("Syntactical Bootstrapping effect size vs. Age (days)") +
  theme(legend.position = "none") 

m_age_young <- rma.mv(d_calc ~ mean_age, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_young)
summary(m_age_young)
## 
## Multivariate Meta-Analysis Model (k = 57; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -115.3058   230.6116   236.6116   242.6336   237.0822   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.2908  0.5393     17     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 55) = 296.5941, p-val < .0001
## 
## Test of Moderators (coefficient 2):
## QM(df = 1) = 0.3081, p-val = 0.5788
## 
## Model Results:
## 
##           estimate      se     zval    pval    ci.lb   ci.ub 
## intrcpt     0.6132  0.3973   1.5433  0.1228  -0.1656  1.3920    
## mean_age   -0.0003  0.0005  -0.5551  0.5788  -0.0012  0.0007    
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
forest(m_age_young,
       header = T,
       slab = ma_data_young$unique_id,
       col = "red",
       cex = .7
)

old, >= 36

ma_data_old %>%
  ggplot(aes(x = age_months, y = d_calc, size = n_1)) +
  geom_point() +
  geom_smooth(method = "lm") +
  geom_smooth(color = "red") +
  ylab("Effect Size") +
  xlab("Age (days)") +
  ggtitle("Syntactical Bootstrapping effect size vs. Age (days)") +
  theme(legend.position = "none") 

m_age_old <- rma.mv(d_calc ~ mean_age, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_old)
summary(m_age_old)
## 
## Multivariate Meta-Analysis Model (k = 22; method: REML)
## 
##   logLik  Deviance       AIC       BIC      AICc 
## -42.8041   85.6082   91.6082   94.5954   93.1082   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.0322  0.1795      6     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 20) = 107.1658, p-val < .0001
## 
## Test of Moderators (coefficient 2):
## QM(df = 1) = 0.9219, p-val = 0.3370
## 
## Model Results:
## 
##           estimate      se     zval    pval    ci.lb   ci.ub 
## intrcpt     0.7414  0.4639   1.5982  0.1100  -0.1679  1.6507    
## mean_age   -0.0003  0.0003  -0.9602  0.3370  -0.0010  0.0003    
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

age and test type

all

ma_data %>% 
  mutate(age_months = mean_age/30.44) %>% 
  ggplot(aes(x = age_months, y = d_calc, size = n_1, color = test_method)) +
  geom_point() +
  geom_smooth(method = "lm") +
  ylab("Effect Size") +
  xlab("Age (days)") +
  ggtitle("Syntactical Bootstrapping effect size vs. Age (days), breakdown by method") 

m_age_method <- rma.mv(d_calc ~ mean_age + test_method, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data)
summary(m_age_method)
## 
## Multivariate Meta-Analysis Model (k = 79; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -158.3123   316.6246   324.6246   333.9475   325.1880   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.2008  0.4481     22     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 76) = 392.3227, p-val < .0001
## 
## Test of Moderators (coefficients 2:3):
## QM(df = 2) = 8.5127, p-val = 0.0142
## 
## Model Results:
## 
##                   estimate      se     zval    pval    ci.lb    ci.ub 
## intrcpt             0.7753  0.1910   4.0591  <.0001   0.4010   1.1497  *** 
## mean_age           -0.0005  0.0002  -2.8250  0.0047  -0.0009  -0.0002   ** 
## test_methodpoint    0.4311  0.2834   1.5213  0.1282  -0.1243   0.9865      
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

age and sentence structure

all

ma_data %>% 
  mutate(age_months = mean_age/30.44) %>% 
  ggplot(aes(x = age_months, y = d_calc, size = n_1, , color = sentence_structure)) +
  geom_point() +
  geom_smooth(method = "lm") +
  ylab("Effect Size") +
  xlab("Age (days)") +
  ggtitle("Syntactical Bootstrapping effect size vs. Age (days)") 

m_age_sentence <- rma.mv(d_calc ~ mean_age + sentence_structure, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data)
summary(m_age_sentence)
## 
## Multivariate Meta-Analysis Model (k = 79; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -152.4985   304.9970   312.9970   322.3200   313.5604   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.2730  0.5225     22     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 76) = 403.2276, p-val < .0001
## 
## Test of Moderators (coefficients 2:3):
## QM(df = 2) = 21.7578, p-val < .0001
## 
## Model Results:
## 
##                               estimate      se     zval    pval    ci.lb 
## intrcpt                         0.4870  0.2162   2.2527  0.0243   0.0633 
## mean_age                       -0.0003  0.0002  -1.8192  0.0689  -0.0007 
## sentence_structuretransitive    0.3208  0.0814   3.9422  <.0001   0.1613 
##                                ci.ub 
## intrcpt                       0.9107    * 
## mean_age                      0.0000    . 
## sentence_structuretransitive  0.4803  *** 
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Interaction:

m_age_sentence <- rma.mv(d_calc ~ mean_age * sentence_structure, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data)
summary(m_age_sentence)
## 
## Multivariate Meta-Analysis Model (k = 79; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -148.9253   297.8506   307.8506   319.4380   308.7201   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.2577  0.5076     22     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 75) = 383.4885, p-val < .0001
## 
## Test of Moderators (coefficients 2:4):
## QM(df = 3) = 29.2563, p-val < .0001
## 
## Model Results:
## 
##                                        estimate      se     zval    pval 
## intrcpt                                  0.7310  0.2312   3.1611  0.0016 
## mean_age                                -0.0006  0.0002  -2.9344  0.0033 
## sentence_structuretransitive            -0.2585  0.2237  -1.1556  0.2478 
## mean_age:sentence_structuretransitive    0.0006  0.0002   2.7696  0.0056 
##                                          ci.lb    ci.ub 
## intrcpt                                 0.2778   1.1842  ** 
## mean_age                               -0.0010  -0.0002  ** 
## sentence_structuretransitive           -0.6968   0.1799     
## mean_age:sentence_structuretransitive   0.0002   0.0011  ** 
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

< 36, young

ma_data_young_only <- ma_data %>%
    mutate(age_months = mean_age/30.44) %>% 
    filter(age_months < 36) 

ma_data_young_only %>% 
  filter(sentence_structure != "bare_verb") %>%
  ggplot(aes(x = age_months, y = d_calc, size = n_1, , color = sentence_structure)) +
  geom_point() +
  geom_smooth(method = "lm") +
  ylab("Effect Size") +
  xlab("Age (days)") +
  ggtitle("Syntactical Bootstrapping effect size vs. Age (days)") 

m_age_sentence_young <- rma.mv(d_calc ~ mean_age + sentence_structure, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_young_only)
summary(m_age_sentence_young)
## 
## Multivariate Meta-Analysis Model (k = 57; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -114.4871   228.9742   236.9742   244.9301   237.7905   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.3256  0.5706     17     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 54) = 293.9865, p-val < .0001
## 
## Test of Moderators (coefficients 2:3):
## QM(df = 2) = 2.2202, p-val = 0.3295
## 
## Model Results:
## 
##                               estimate      se     zval    pval    ci.lb 
## intrcpt                         0.4825  0.4118   1.1716  0.2413  -0.3247 
## mean_age                       -0.0002  0.0005  -0.3992  0.6898  -0.0012 
## sentence_structuretransitive    0.1316  0.0946   1.3915  0.1641  -0.0538 
##                                ci.ub 
## intrcpt                       1.2897    
## mean_age                      0.0008    
## sentence_structuretransitive  0.3169    
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

interaction:

m_age_sentence_young <- rma.mv(d_calc ~ mean_age * sentence_structure, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_young_only)
summary(m_age_sentence_young)
## 
## Multivariate Meta-Analysis Model (k = 57; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -111.3248   222.6496   232.6496   242.5010   233.9262   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.3002  0.5479     17     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 53) = 281.8953, p-val < .0001
## 
## Test of Moderators (coefficients 2:4):
## QM(df = 3) = 8.8233, p-val = 0.0317
## 
## Model Results:
## 
##                                        estimate      se     zval    pval 
## intrcpt                                  0.1283  0.4318   0.2971  0.7664 
## mean_age                                 0.0003  0.0005   0.6080  0.5432 
## sentence_structuretransitive             1.3145  0.4685   2.8058  0.0050 
## mean_age:sentence_structuretransitive   -0.0016  0.0006  -2.5857  0.0097 
##                                          ci.lb    ci.ub 
## intrcpt                                -0.7181   0.9747     
## mean_age                               -0.0007   0.0014     
## sentence_structuretransitive            0.3963   2.2328  ** 
## mean_age:sentence_structuretransitive  -0.0029  -0.0004  ** 
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

>= 36, old

ma_data_old <- ma_data %>%
    mutate(age_months = mean_age/30.44) %>% 
    filter(age_months > 36 | age_months == 36) 

ma_data_old %>% 
  filter(sentence_structure != "bare_verb") %>%
  ggplot(aes(x = age_months, y = d_calc, size = n_1, color = sentence_structure)) +
  geom_point() +
  geom_smooth(method = "lm") +
  ylab("Effect Size") +
  xlab("Age (days)") +
  ggtitle("Syntactical Bootstrapping effect size vs. Age (days)") 

m_age_sentence_old <- rma.mv(d_calc ~ mean_age + sentence_structure, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_old)
summary(m_age_sentence_old)
## 
## Multivariate Meta-Analysis Model (k = 22; method: REML)
## 
##   logLik  Deviance       AIC       BIC      AICc 
## -27.4835   54.9669   62.9669   66.7447   65.8240   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.0000  0.0000      6     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 19) = 71.5803, p-val < .0001
## 
## Test of Moderators (coefficients 2:3):
## QM(df = 2) = 35.7112, p-val < .0001
## 
## Model Results:
## 
##                               estimate      se     zval    pval    ci.lb 
## intrcpt                         0.1217  0.4116   0.2958  0.7674  -0.6849 
## mean_age                       -0.0002  0.0003  -0.5996  0.5488  -0.0008 
## sentence_structuretransitive    0.6793  0.1139   5.9654  <.0001   0.4561 
##                                ci.ub 
## intrcpt                       0.9284      
## mean_age                      0.0004      
## sentence_structuretransitive  0.9025  *** 
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

interaction

m_age_sentence_old <- rma.mv(d_calc ~ mean_age * sentence_structure, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_old)
summary(m_age_sentence_old)
## 
## Multivariate Meta-Analysis Model (k = 22; method: REML)
## 
##   logLik  Deviance       AIC       BIC      AICc 
## -27.3891   54.7783   64.7783   69.2301   69.7783   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.0000  0.0000      6     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 18) = 70.3088, p-val < .0001
## 
## Test of Moderators (coefficients 2:4):
## QM(df = 3) = 36.9827, p-val < .0001
## 
## Model Results:
## 
##                                        estimate      se     zval    pval 
## intrcpt                                  0.9460  0.8389   1.1277  0.2595 
## mean_age                                -0.0008  0.0006  -1.2770  0.2016 
## sentence_structuretransitive            -0.3953  0.9598  -0.4119  0.6804 
## mean_age:sentence_structuretransitive    0.0008  0.0007   1.1276  0.2595 
##                                          ci.lb   ci.ub 
## intrcpt                                -0.6982  2.5902    
## mean_age                               -0.0020  0.0004    
## sentence_structuretransitive           -2.2765  1.4858    
## mean_age:sentence_structuretransitive  -0.0006  0.0022    
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

age with stimuli_actor

all

ma_data %>% 
  mutate(age_months = mean_age/30.44) %>% 
  ggplot(aes(x = age_months, y = d_calc, size = n_1, color = stimuli_actor)) +
  geom_point() +
  geom_smooth(method = "lm") +
  ylab("Effect Size") +
  xlab("Age (days)") +
  ggtitle("Syntactical Bootstrapping effect size vs. Age (days), breakdown by stimuli_actor") 

m_age_stimuli_actor <- rma.mv(d_calc ~ mean_age + stimuli_actor, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data)
m_age_stimuli_actor_interaction <- rma.mv(d_calc ~ mean_age * stimuli_actor, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data)
summary(m_age_stimuli_actor)
## 
## Multivariate Meta-Analysis Model (k = 79; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -159.2579   318.5159   326.5159   335.8388   327.0793   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.2247  0.4740     22     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 76) = 402.2188, p-val < .0001
## 
## Test of Moderators (coefficients 2:3):
## QM(df = 2) = 6.5734, p-val = 0.0374
## 
## Model Results:
## 
##                      estimate      se     zval    pval    ci.lb   ci.ub 
## intrcpt                0.7731  0.1958   3.9484  <.0001   0.3893  1.1568  *** 
## mean_age              -0.0004  0.0002  -1.6743  0.0941  -0.0008  0.0001    . 
## stimuli_actorperson   -0.0975  0.1612  -0.6051  0.5451  -0.4135  0.2184      
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
summary(m_age_stimuli_actor_interaction)
## 
## Multivariate Meta-Analysis Model (k = 79; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -158.3652   316.7303   326.7303   338.3178   327.5999   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.2204  0.4695     22     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 75) = 393.4459, p-val < .0001
## 
## Test of Moderators (coefficients 2:4):
## QM(df = 3) = 7.0701, p-val = 0.0697
## 
## Model Results:
## 
##                               estimate      se     zval    pval    ci.lb 
## intrcpt                         0.6344  0.2769   2.2913  0.0219   0.0917 
## mean_age                       -0.0002  0.0003  -0.7799  0.4354  -0.0008 
## stimuli_actorperson             0.2888  0.5721   0.5048  0.6137  -0.8325 
## mean_age:stimuli_actorperson   -0.0004  0.0006  -0.7034  0.4818  -0.0016 
##                                ci.ub 
## intrcpt                       1.1771  * 
## mean_age                      0.0003    
## stimuli_actorperson           1.4101    
## mean_age:stimuli_actorperson  0.0007    
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

<36, young

ma_data_young_only %>% 
  mutate(age_months = mean_age/30.44) %>% 
  ggplot(aes(x = age_months, y = d_calc, size = n_1, color = stimuli_actor)) +
  geom_point() +
  geom_smooth(method = "lm") +
  ylab("Effect Size") +
  xlab("Age (days)") +
  ggtitle("Syntactical Bootstrapping effect size vs. Age (days), breakdown by stimuli_actor, young only") 

m_age_stimuli_actor_young <- rma.mv(d_calc ~ mean_age + stimuli_actor, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_young_only)
m_age_stimuli_actor_interaction_young <- rma.mv(d_calc ~ mean_age * stimuli_actor, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_young_only)
summary(m_age_stimuli_actor_young)
## 
## Multivariate Meta-Analysis Model (k = 57; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -114.4670   228.9340   236.9340   244.8899   237.7503   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.3085  0.5554     17     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 54) = 296.5893, p-val < .0001
## 
## Test of Moderators (coefficients 2:3):
## QM(df = 2) = 0.3155, p-val = 0.8541
## 
## Model Results:
## 
##                      estimate      se     zval    pval    ci.lb   ci.ub 
## intrcpt                0.6312  0.4263   1.4808  0.1387  -0.2042  1.4667    
## mean_age              -0.0003  0.0006  -0.5047  0.6138  -0.0016  0.0009    
## stimuli_actorperson    0.0387  0.2706   0.1429  0.8864  -0.4918  0.5691    
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
summary(m_age_stimuli_actor_interaction_young)
## 
## Multivariate Meta-Analysis Model (k = 57; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -113.4216   226.8432   236.8432   246.6946   238.1198   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.3492  0.5910     17     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 53) = 290.5912, p-val < .0001
## 
## Test of Moderators (coefficients 2:4):
## QM(df = 3) = 0.6425, p-val = 0.8866
## 
## Model Results:
## 
##                               estimate      se     zval    pval    ci.lb 
## intrcpt                         1.0220  0.7973   1.2819  0.1999  -0.5407 
## mean_age                       -0.0008  0.0011  -0.7531  0.4514  -0.0030 
## stimuli_actorperson            -0.8348  1.4629  -0.5706  0.5683  -3.7021 
## mean_age:stimuli_actorperson    0.0011  0.0019   0.6016  0.5474  -0.0026 
##                                ci.ub 
## intrcpt                       2.5847    
## mean_age                      0.0013    
## stimuli_actorperson           2.0325    
## mean_age:stimuli_actorperson  0.0048    
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

>= 36, old

ma_data_old %>% 
  mutate(age_months = mean_age/30.44) %>% 
  ggplot(aes(x = age_months, y = d_calc, size = n_1, color = stimuli_actor)) +
  geom_point() +
  ylab("Effect Size") +
  xlab("Age (days)") +
  ggtitle("Syntactical Bootstrapping effect size vs. Age (days), breakdown by stimuli_actor, old only") 

m_age_stimuli_actor_old <- rma.mv(d_calc ~ mean_age + stimuli_actor, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_old)
m_age_stimuli_actor_interaction_old <- rma.mv(d_calc ~ mean_age * stimuli_actor, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_old)
summary(m_age_stimuli_actor_old)
## 
## Multivariate Meta-Analysis Model (k = 22; method: REML)
## 
##   logLik  Deviance       AIC       BIC      AICc 
## -39.5886   79.1771   87.1771   90.9549   90.0343   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.0000  0.0000      6     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 19) = 95.8680, p-val < .0001
## 
## Test of Moderators (coefficients 2:3):
## QM(df = 2) = 11.4235, p-val = 0.0033
## 
## Model Results:
## 
##                      estimate      se     zval    pval    ci.lb    ci.ub 
## intrcpt                0.8306  0.4191   1.9818  0.0475   0.0092   1.6520    * 
## mean_age              -0.0003  0.0003  -0.9079  0.3639  -0.0009   0.0003      
## stimuli_actorperson   -0.4021  0.1196  -3.3612  0.0008  -0.6366  -0.1676  *** 
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
summary(m_age_stimuli_actor_interaction_old)
## 
## Multivariate Meta-Analysis Model (k = 22; method: REML)
## 
##   logLik  Deviance       AIC       BIC      AICc 
## -39.9337   79.8673   89.8673   94.3192   94.8673   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.0000  0.0000      6     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 18) = 95.5949, p-val < .0001
## 
## Test of Moderators (coefficients 2:4):
## QM(df = 3) = 11.6966, p-val = 0.0085
## 
## Model Results:
## 
##                               estimate      se     zval    pval    ci.lb 
## intrcpt                         0.8938  0.4362   2.0490  0.0405   0.0388 
## mean_age                       -0.0003  0.0003  -1.0180  0.3087  -0.0009 
## stimuli_actorperson            -1.1721  1.4782  -0.7929  0.4278  -4.0693 
## mean_age:stimuli_actorperson    0.0006  0.0011   0.5226  0.6013  -0.0016 
##                                ci.ub 
## intrcpt                       1.7487  * 
## mean_age                      0.0003    
## stimuli_actorperson           1.7251    
## mean_age:stimuli_actorperson  0.0028    
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

age with presentation_type

Decided to only compare asynchronous and immediate after:

ma_data %>% 
  count(presentation_type)
## # A tibble: 3 x 2
##   presentation_type     n
##   <chr>             <int>
## 1 asynchronous         35
## 2 immediate_after      32
## 3 simultaneous         12
ma_data_pt <- ma_data %>% 
  filter(presentation_type != "simultaneous") 

ma_data_pt_young <- ma_data_pt %>% 
  mutate(age_months = mean_age/30.44) %>% 
    filter(age_months < 36) 


ma_data_pt_old <- ma_data_pt %>% 
  mutate(age_months = mean_age/30.44) %>% 
    filter(age_months > 36 | age_months == 36) 

all

ma_data_pt %>% 
  mutate(age_months = mean_age/30.44) %>% 
  ggplot(aes(x = age_months, y = d_calc, size = n_1,  color = presentation_type)) +
  geom_point() +
  geom_smooth(method = "lm") +
  ylab("Effect Size") +
  xlab("Age (days)") +
  ggtitle("Syntactical Bootstrapping effect size vs. Age (days), breakdown by presentation type") 

m_age_pt <- rma.mv(d_calc ~ mean_age + presentation_type, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_pt)

m_age_pt_interaction <- rma.mv(d_calc ~ mean_age * presentation_type, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_pt)

summary(m_age_pt)
## 
## Multivariate Meta-Analysis Model (k = 67; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -117.7657   235.5315   243.5315   252.1670   244.2095   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.3222  0.5676     14     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 64) = 309.1098, p-val < .0001
## 
## Test of Moderators (coefficients 2:3):
## QM(df = 2) = 6.5376, p-val = 0.0381
## 
## Model Results:
## 
##                                   estimate      se     zval    pval    ci.lb 
## intrcpt                             0.8560  0.2336   3.6648  0.0002   0.3982 
## mean_age                           -0.0005  0.0002  -2.5538  0.0107  -0.0008 
## presentation_typeimmediate_after    0.0166  0.1889   0.0880  0.9299  -0.3537 
##                                     ci.ub 
## intrcpt                            1.3138  *** 
## mean_age                          -0.0001    * 
## presentation_typeimmediate_after   0.3869      
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
summary(m_age_pt_interaction)
## 
## Multivariate Meta-Analysis Model (k = 67; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -117.0993   234.1987   244.1987   254.9144   245.2513   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.3151  0.5613     14     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 63) = 303.2844, p-val < .0001
## 
## Test of Moderators (coefficients 2:4):
## QM(df = 3) = 7.6700, p-val = 0.0533
## 
## Model Results:
## 
##                                            estimate      se     zval    pval 
## intrcpt                                      0.9282  0.2424   3.8293  0.0001 
## mean_age                                    -0.0006  0.0002  -2.7277  0.0064 
## presentation_typeimmediate_after            -0.2834  0.3412  -0.8305  0.4062 
## mean_age:presentation_typeimmediate_after    0.0004  0.0004   1.0605  0.2889 
##                                              ci.lb    ci.ub 
## intrcpt                                     0.4531   1.4033  *** 
## mean_age                                   -0.0010  -0.0002   ** 
## presentation_typeimmediate_after           -0.9522   0.3854      
## mean_age:presentation_typeimmediate_after  -0.0004   0.0012      
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

<36, young

ma_data_pt_young %>% 
  mutate(age_months = mean_age/30.44) %>% 
  ggplot(aes(x = age_months, y = d_calc, size = n_1,  color = presentation_type)) +
  geom_point() +
  geom_smooth(method = "lm") +
  ylab("Effect Size") +
  xlab("Age (days)") +
  ggtitle("Syntactical Bootstrapping effect size vs. Age (days), breakdown by presentation type, young only") 

m_data_pt_young <- rma.mv(d_calc ~ mean_age + presentation_type, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_pt_young)

m_age_pt_interaction_young <- rma.mv(d_calc ~ mean_age * presentation_type, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_pt_young)

summary(m_age_pt)
## 
## Multivariate Meta-Analysis Model (k = 67; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -117.7657   235.5315   243.5315   252.1670   244.2095   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.3222  0.5676     14     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 64) = 309.1098, p-val < .0001
## 
## Test of Moderators (coefficients 2:3):
## QM(df = 2) = 6.5376, p-val = 0.0381
## 
## Model Results:
## 
##                                   estimate      se     zval    pval    ci.lb 
## intrcpt                             0.8560  0.2336   3.6648  0.0002   0.3982 
## mean_age                           -0.0005  0.0002  -2.5538  0.0107  -0.0008 
## presentation_typeimmediate_after    0.0166  0.1889   0.0880  0.9299  -0.3537 
##                                     ci.ub 
## intrcpt                            1.3138  *** 
## mean_age                          -0.0001    * 
## presentation_typeimmediate_after   0.3869      
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
summary(m_age_pt_interaction)
## 
## Multivariate Meta-Analysis Model (k = 67; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -117.0993   234.1987   244.1987   254.9144   245.2513   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.3151  0.5613     14     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 63) = 303.2844, p-val < .0001
## 
## Test of Moderators (coefficients 2:4):
## QM(df = 3) = 7.6700, p-val = 0.0533
## 
## Model Results:
## 
##                                            estimate      se     zval    pval 
## intrcpt                                      0.9282  0.2424   3.8293  0.0001 
## mean_age                                    -0.0006  0.0002  -2.7277  0.0064 
## presentation_typeimmediate_after            -0.2834  0.3412  -0.8305  0.4062 
## mean_age:presentation_typeimmediate_after    0.0004  0.0004   1.0605  0.2889 
##                                              ci.lb    ci.ub 
## intrcpt                                     0.4531   1.4033  *** 
## mean_age                                   -0.0010  -0.0002   ** 
## presentation_typeimmediate_after           -0.9522   0.3854      
## mean_age:presentation_typeimmediate_after  -0.0004   0.0012      
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

>= 36, old

ma_data_pt_old %>% 
  mutate(age_months = mean_age/30.44) %>% 
  ggplot(aes(x = age_months, y = d_calc, size = n_1,  color = presentation_type)) +
  geom_point() +
  geom_smooth(method = "lm") +
  ylab("Effect Size") +
  xlab("Age (days)") +
  ggtitle("Syntactical Bootstrapping effect size vs. Age (days), breakdown by presentation type, young only") 

age with character_identification

all

ma_data %>% 
  mutate(age_months = mean_age/30.44) %>% 
  ggplot(aes(x = age_months, y = d_calc, size = n_1, color = character_identification)) +
  geom_point() +
  geom_smooth(method = "lm") +
  ylab("Effect Size") +
  xlab("Age (days)") +
  ggtitle("Syntactical Bootstrapping effect size vs. Age (days), breakdown by character_identification") 

m_age_ci <- rma.mv(d_calc ~ mean_age + character_identification, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data)
m_age_stimuli_ci_interaction <- rma.mv(d_calc ~ mean_age * character_identification, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data)
summary(m_age_ci)
## 
## Multivariate Meta-Analysis Model (k = 79; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -159.0126   318.0251   326.0251   335.3481   326.5885   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.2252  0.4745     22     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 76) = 401.9504, p-val < .0001
## 
## Test of Moderators (coefficients 2:3):
## QM(df = 2) = 6.6079, p-val = 0.0367
## 
## Model Results:
## 
##                              estimate      se     zval    pval    ci.lb 
## intrcpt                        0.7360  0.2098   3.5078  0.0005   0.3247 
## mean_age                      -0.0005  0.0002  -2.5629  0.0104  -0.0008 
## character_identificationyes    0.1418  0.2241   0.6328  0.5269  -0.2975 
##                                ci.ub 
## intrcpt                       1.1472  *** 
## mean_age                     -0.0001    * 
## character_identificationyes   0.5811      
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
summary(m_age_stimuli_ci_interaction)
## 
## Multivariate Meta-Analysis Model (k = 79; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -158.5201   317.0401   327.0401   338.6276   327.9097   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.2271  0.4765     22     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 75) = 396.8563, p-val < .0001
## 
## Test of Moderators (coefficients 2:4):
## QM(df = 3) = 7.3253, p-val = 0.0622
## 
## Model Results:
## 
##                                       estimate      se     zval    pval 
## intrcpt                                 0.8207  0.2326   3.5280  0.0004 
## mean_age                               -0.0006  0.0002  -2.5931  0.0095 
## character_identificationyes            -0.1696  0.4303  -0.3942  0.6934 
## mean_age:character_identificationyes    0.0003  0.0004   0.8487  0.3960 
##                                         ci.lb    ci.ub 
## intrcpt                                0.3648   1.2767  *** 
## mean_age                              -0.0010  -0.0001   ** 
## character_identificationyes           -1.0130   0.6738      
## mean_age:character_identificationyes  -0.0004   0.0011      
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

<36, young

ma_data_young_only %>% 
  mutate(age_months = mean_age/30.44) %>%
  ggplot(aes(x = age_months, y = d_calc, size = n_1, color = character_identification)) +
  geom_point() +
  geom_smooth(method = "lm") +
  ylab("Effect Size") +
  xlab("Age (days)") +
  ggtitle("Syntactical Bootstrapping effect size vs. Age (days), breakdown by character_identification, young only") 

m_age_ci_young <- rma.mv(d_calc ~ mean_age + character_identification, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_young_only)
m_age_ci_interaction_young <- rma.mv(d_calc ~ mean_age * character_identification, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_young_only)
summary(m_age_ci_young)
## 
## Multivariate Meta-Analysis Model (k = 57; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -114.3668   228.7337   236.7337   244.6896   237.5500   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.3144  0.5607     17     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 54) = 294.3994, p-val < .0001
## 
## Test of Moderators (coefficients 2:3):
## QM(df = 2) = 0.2973, p-val = 0.8619
## 
## Model Results:
## 
##                              estimate      se     zval    pval    ci.lb   ci.ub 
## intrcpt                        0.6038  0.4060   1.4874  0.1369  -0.1919  1.3995 
## mean_age                      -0.0003  0.0005  -0.5450  0.5857  -0.0012  0.0007 
## character_identificationyes    0.0241  0.3052   0.0791  0.9369  -0.5740  0.6223 
##  
## intrcpt 
## mean_age 
## character_identificationyes 
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
summary(m_age_ci_interaction_young)
## 
## Multivariate Meta-Analysis Model (k = 57; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -113.5404   227.0809   237.0809   246.9323   238.3575   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.3441  0.5866     17     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 53) = 291.6547, p-val < .0001
## 
## Test of Moderators (coefficients 2:4):
## QM(df = 3) = 0.6826, p-val = 0.8773
## 
## Model Results:
## 
##                                       estimate      se     zval    pval 
## intrcpt                                 0.6697  0.4253   1.5747  0.1153 
## mean_age                               -0.0004  0.0005  -0.6893  0.4906 
## character_identificationyes            -0.8853  1.4630  -0.6051  0.5451 
## mean_age:character_identificationyes    0.0011  0.0018   0.6359  0.5249 
##                                         ci.lb   ci.ub 
## intrcpt                               -0.1638  1.5032    
## mean_age                              -0.0014  0.0007    
## character_identificationyes           -3.7528  1.9823    
## mean_age:character_identificationyes  -0.0023  0.0046    
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

>= 36, old

ma_data_old %>% 
  mutate(age_months = mean_age/30.44) %>%
  ggplot(aes(x = age_months, y = d_calc, size = n_1, color = character_identification)) +
  geom_point() +
  ylab("Effect Size") +
  xlab("Age (days)") +
  ggtitle("Syntactical Bootstrapping effect size vs. Age (days), breakdown by character_identification, old only") 

age with practice_phase

all

ma_data %>% 
  mutate(age_months = mean_age/30.44) %>% 
  ggplot(aes(x = age_months, y = d_calc, size = n_1, color = practice_phase)) +
  geom_point() +
  geom_smooth(method = "lm") +
  ylab("Effect Size") +
  xlab("Age (days)") +
  ggtitle("Syntactical Bootstrapping effect size vs. Age (days), breakdown by practice_phase") 

m_age_pf <- rma.mv(d_calc ~ mean_age + character_identification, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data)
m_age_pf_interaction <- rma.mv(d_calc ~ mean_age * character_identification, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data)
summary(m_age_ci)
## 
## Multivariate Meta-Analysis Model (k = 79; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -159.0126   318.0251   326.0251   335.3481   326.5885   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.2252  0.4745     22     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 76) = 401.9504, p-val < .0001
## 
## Test of Moderators (coefficients 2:3):
## QM(df = 2) = 6.6079, p-val = 0.0367
## 
## Model Results:
## 
##                              estimate      se     zval    pval    ci.lb 
## intrcpt                        0.7360  0.2098   3.5078  0.0005   0.3247 
## mean_age                      -0.0005  0.0002  -2.5629  0.0104  -0.0008 
## character_identificationyes    0.1418  0.2241   0.6328  0.5269  -0.2975 
##                                ci.ub 
## intrcpt                       1.1472  *** 
## mean_age                     -0.0001    * 
## character_identificationyes   0.5811      
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
summary(m_age_stimuli_ci_interaction)
## 
## Multivariate Meta-Analysis Model (k = 79; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -158.5201   317.0401   327.0401   338.6276   327.9097   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.2271  0.4765     22     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 75) = 396.8563, p-val < .0001
## 
## Test of Moderators (coefficients 2:4):
## QM(df = 3) = 7.3253, p-val = 0.0622
## 
## Model Results:
## 
##                                       estimate      se     zval    pval 
## intrcpt                                 0.8207  0.2326   3.5280  0.0004 
## mean_age                               -0.0006  0.0002  -2.5931  0.0095 
## character_identificationyes            -0.1696  0.4303  -0.3942  0.6934 
## mean_age:character_identificationyes    0.0003  0.0004   0.8487  0.3960 
##                                         ci.lb    ci.ub 
## intrcpt                                0.3648   1.2767  *** 
## mean_age                              -0.0010  -0.0001   ** 
## character_identificationyes           -1.0130   0.6738      
## mean_age:character_identificationyes  -0.0004   0.0011      
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

<36, young

ma_data_young_only %>% 
  mutate(age_months = mean_age/30.44) %>%
  ggplot(aes(x = age_months, y = d_calc, size = n_1, color = practice_phase)) +
  geom_point() +
  geom_smooth(method = "lm") +
  ylab("Effect Size") +
  xlab("Age (days)") +
  ggtitle("Syntactical Bootstrapping effect size vs. Age (days), breakdown by practice_phase, young only") 

m_age_pf_young <- rma.mv(d_calc ~ mean_age + practice_phase, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_young_only)
m_age_pf_interaction_young <- rma.mv(d_calc ~ mean_age * practice_phase, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_young_only)
summary(m_age_pf_young)
## 
## Multivariate Meta-Analysis Model (k = 57; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -114.5234   229.0468   237.0468   245.0027   237.8631   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.2749  0.5243     17     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 54) = 290.5875, p-val < .0001
## 
## Test of Moderators (coefficients 2:3):
## QM(df = 2) = 0.8004, p-val = 0.6702
## 
## Model Results:
## 
##                    estimate      se     zval    pval    ci.lb   ci.ub 
## intrcpt              0.7492  0.4393   1.7055  0.0881  -0.1118  1.6103  . 
## mean_age            -0.0005  0.0006  -0.8686  0.3851  -0.0017  0.0007    
## practice_phaseyes    0.1335  0.1928   0.6921  0.4889  -0.2445  0.5114    
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
summary(m_age_pf_interaction_young)
## 
## Multivariate Meta-Analysis Model (k = 57; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -113.3324   226.6648   236.6648   246.5162   237.9414   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.3045  0.5518     17     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 53) = 288.4535, p-val < .0001
## 
## Test of Moderators (coefficients 2:4):
## QM(df = 3) = 1.9467, p-val = 0.5835
## 
## Model Results:
## 
##                             estimate      se     zval    pval    ci.lb   ci.ub 
## intrcpt                       1.3243  0.6888   1.9225  0.0545  -0.0258  2.6743 
## mean_age                     -0.0014  0.0010  -1.3847  0.1662  -0.0033  0.0006 
## practice_phaseyes            -0.9787  0.9969  -0.9817  0.3263  -2.9326  0.9753 
## mean_age:practice_phaseyes    0.0016  0.0014   1.1256  0.2603  -0.0012  0.0044 
##  
## intrcpt                     . 
## mean_age 
## practice_phaseyes 
## mean_age:practice_phaseyes 
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

>= 36, old

ma_data_old %>% 
  mutate(age_months = mean_age/30.44) %>%
  ggplot(aes(x = age_months, y = d_calc, size = n_1, color = practice_phase)) +
  geom_point() +
  ylab("Effect Size") +
  xlab("Age (days)") +
  ggtitle("Syntactical Bootstrapping effect size vs. Age (days), breakdown by practice_phase, old only") 

age with test_mass_or_distributed

all

ma_data %>% 
  mutate(age_months = mean_age/30.44) %>% 
  ggplot(aes(x = age_months, y = d_calc, size = n_1, color = test_mass_or_distributed)) +
  geom_point() +
  geom_smooth(method = "lm") +
  ylab("Effect Size") +
  xlab("Age (days)") +
  ggtitle("Syntactical Bootstrapping effect size vs. Age (days), breakdown by test type") 

m_age_tt <- rma.mv(d_calc ~ mean_age + test_mass_or_distributed, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data)
m_age_tt_interaction <- rma.mv(d_calc ~ mean_age * test_mass_or_distributed, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data)
summary(m_age_tt)
## 
## Multivariate Meta-Analysis Model (k = 79; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -159.0278   318.0556   326.0556   335.3786   326.6190   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.2317  0.4813     22     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 76) = 407.4472, p-val < .0001
## 
## Test of Moderators (coefficients 2:3):
## QM(df = 2) = 6.2337, p-val = 0.0443
## 
## Model Results:
## 
##                               estimate      se     zval    pval    ci.lb 
## intrcpt                         0.7975  0.2088   3.8202  0.0001   0.3883 
## mean_age                       -0.0004  0.0002  -2.4952  0.0126  -0.0008 
## test_mass_or_distributedmass   -0.0393  0.2515  -0.1562  0.8759  -0.5323 
##                                 ci.ub 
## intrcpt                        1.2067  *** 
## mean_age                      -0.0001    * 
## test_mass_or_distributedmass   0.4537      
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
summary(m_age_tt_interaction)
## 
## Multivariate Meta-Analysis Model (k = 79; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -158.2020   316.4039   326.4039   337.9913   327.2735   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.2271  0.4766     22     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 75) = 397.5932, p-val < .0001
## 
## Test of Moderators (coefficients 2:4):
## QM(df = 3) = 7.4416, p-val = 0.0591
## 
## Model Results:
## 
##                                        estimate      se     zval    pval 
## intrcpt                                  0.5393  0.3133   1.7215  0.0852 
## mean_age                                -0.0002  0.0003  -0.5200  0.6031 
## test_mass_or_distributedmass             0.3369  0.4237   0.7951  0.4266 
## mean_age:test_mass_or_distributedmass   -0.0004  0.0004  -1.0993  0.2717 
##                                          ci.lb   ci.ub 
## intrcpt                                -0.0747  1.1533  . 
## mean_age                               -0.0008  0.0004    
## test_mass_or_distributedmass           -0.4936  1.1673    
## mean_age:test_mass_or_distributedmass  -0.0012  0.0003    
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

<36, young

ma_data_young_only %>% 
  mutate(age_months = mean_age/30.44) %>%
  ggplot(aes(x = age_months, y = d_calc, size = n_1, color = test_mass_or_distributed)) +
  geom_point() +
  geom_smooth(method = "lm") +
  ylab("Effect Size") +
  xlab("Age (days)") +
  ggtitle("Syntactical Bootstrapping effect size vs. Age (days), breakdown by test_mass_or_distributed, young only") 

m_age_tt_young <- rma.mv(d_calc ~ mean_age + test_mass_or_distributed, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_young_only)
m_age_tt_interaction_young <- rma.mv(d_calc ~ mean_age * test_mass_or_distributed, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_young_only)
summary(m_age_tt_young)
## 
## Multivariate Meta-Analysis Model (k = 57; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -114.3078   228.6157   236.6157   244.5716   237.4320   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.3143  0.5607     17     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 54) = 296.5626, p-val < .0001
## 
## Test of Moderators (coefficients 2:3):
## QM(df = 2) = 0.3138, p-val = 0.8548
## 
## Model Results:
## 
##                               estimate      se     zval    pval    ci.lb 
## intrcpt                         0.5878  0.4247   1.3838  0.1664  -0.2447 
## mean_age                       -0.0003  0.0005  -0.5162  0.6057  -0.0012 
## test_mass_or_distributedmass    0.0487  0.3230   0.1508  0.8801  -0.5843 
##                                ci.ub 
## intrcpt                       1.4203    
## mean_age                      0.0007    
## test_mass_or_distributedmass  0.6817    
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
summary(m_age_tt_interaction_young)
## 
## Multivariate Meta-Analysis Model (k = 57; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -113.2307   226.4615   236.4615   246.3130   237.7381   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.3410  0.5840     17     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 53) = 296.3776, p-val < .0001
## 
## Test of Moderators (coefficients 2:4):
## QM(df = 3) = 1.3688, p-val = 0.7129
## 
## Model Results:
## 
##                                        estimate      se     zval    pval 
## intrcpt                                 -0.5465  1.1727  -0.4660  0.6412 
## mean_age                                 0.0012  0.0015   0.8064  0.4200 
## test_mass_or_distributedmass             1.3089  1.2612   1.0378  0.2993 
## mean_age:test_mass_or_distributedmass   -0.0016  0.0016  -1.0351  0.3006 
##                                          ci.lb   ci.ub 
## intrcpt                                -2.8450  1.7521    
## mean_age                               -0.0017  0.0040    
## test_mass_or_distributedmass           -1.1629  3.7807    
## mean_age:test_mass_or_distributedmass  -0.0046  0.0014    
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

>= 36, old

ma_data_old %>% 
  mutate(age_months = mean_age/30.44) %>%
  ggplot(aes(x = age_months, y = d_calc, size = n_1, color = test_mass_or_distributed)) +
  geom_point() +
  ylab("Effect Size") +
  xlab("Age (days)") +
  ggtitle("Syntactical Bootstrapping effect size vs. Age (days), breakdown by test_mass_or_distributed, old only") 

repetition only

repetition only

ma_data %>% 
  ggplot(mapping = aes(x = n_repetitions_sentence, y = d_calc)) +
  geom_point() +
  geom_smooth(method = "lm") +
  ylab("Effect Size") +
  xlab("number of repetition per novel verb") +
  ggtitle("Syntactical Bootstrapping effect size vs. number of repetitions for verb") +
  theme_classic() +
  theme(legend.position = "none")

m_rep <- rma.mv(d_calc ~ n_repetitions_sentence , V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data)
summary(m_rep)
## 
## Multivariate Meta-Analysis Model (k = 79; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -162.0286   324.0571   330.0571   337.0886   330.3859   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.2328  0.4825     22     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 77) = 409.4680, p-val < .0001
## 
## Test of Moderators (coefficient 2):
## QM(df = 1) = 1.6326, p-val = 0.2013
## 
## Model Results:
## 
##                         estimate      se    zval    pval    ci.lb   ci.ub 
## intrcpt                   0.2566  0.1504  1.7059  0.0880  -0.0382  0.5514  . 
## n_repetitions_sentence    0.0156  0.0122  1.2777  0.2013  -0.0083  0.0395    
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

repetition only, young

ma_data_young %>% 
  ggplot(mapping = aes(x = n_repetitions_sentence, y = d_calc)) +
  geom_point() +
  geom_smooth(method = "lm") +
  ylab("Effect Size") +
  xlab("number of repetition per novel verb") +
  ggtitle("Syntactical Bootstrapping effect size vs. number of repetitions for verb") +
  theme_classic() +
  theme(legend.position = "none")

m_rep_young <- rma.mv(d_calc ~ n_repetitions_sentence , V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_young)
summary(m_rep_young)
## 
## Multivariate Meta-Analysis Model (k = 57; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -115.0254   230.0508   236.0508   242.0727   236.5213   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.3061  0.5533     17     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 55) = 301.9470, p-val < .0001
## 
## Test of Moderators (coefficient 2):
## QM(df = 1) = 0.4143, p-val = 0.5198
## 
## Model Results:
## 
##                         estimate      se    zval    pval    ci.lb   ci.ub 
## intrcpt                   0.3186  0.1995  1.5974  0.1102  -0.0723  0.7095    
## n_repetitions_sentence    0.0090  0.0140  0.6437  0.5198  -0.0184  0.0365    
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

repetition only, old

ma_data_old %>% 
  ggplot(mapping = aes(x = n_repetitions_sentence, y = d_calc)) +
  geom_point() +
  geom_smooth(method = "lm") +
  ylab("Effect Size") +
  xlab("number of repetition per novel verb") +
  ggtitle("Syntactical Bootstrapping effect size vs. number of repetitions for verb") +
  theme_classic() +
  theme(legend.position = "none")

m_rep_old <- rma.mv(d_calc ~ n_repetitions_sentence , V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_old)
summary(m_rep_old)
## 
## Multivariate Meta-Analysis Model (k = 22; method: REML)
## 
##   logLik  Deviance       AIC       BIC      AICc 
## -41.3794   82.7587   88.7587   91.7459   90.2587   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.0000  0.0000      6     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 20) = 100.3330, p-val < .0001
## 
## Test of Moderators (coefficient 2):
## QM(df = 1) = 6.9585, p-val = 0.0083
## 
## Model Results:
## 
##                         estimate      se     zval    pval    ci.lb    ci.ub 
## intrcpt                   0.5495  0.0923   5.9557  <.0001   0.3687   0.7304 
## n_repetitions_sentence   -0.0460  0.0175  -2.6379  0.0083  -0.0802  -0.0118 
##  
## intrcpt                 *** 
## n_repetitions_sentence   ** 
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

repetition with age

correlation

all

m_rep_age <- rma.mv(d_calc ~ mean_age + n_repetitions_sentence , V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data)
m_rep_age_interaction <- rma.mv(d_calc ~ mean_age*n_repetitions_sentence , V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data)
summary(m_rep_age)
## 
## Multivariate Meta-Analysis Model (k = 79; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -159.4956   318.9913   326.9913   336.3142   327.5546   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.2270  0.4764     22     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 76) = 407.5030, p-val < .0001
## 
## Test of Moderators (coefficients 2:3):
## QM(df = 2) = 6.2894, p-val = 0.0431
## 
## Model Results:
## 
##                         estimate      se     zval    pval    ci.lb    ci.ub 
## intrcpt                   0.7339  0.2660   2.7585  0.0058   0.2124   1.2553  ** 
## mean_age                 -0.0004  0.0002  -2.1621  0.0306  -0.0008  -0.0000   * 
## n_repetitions_sentence    0.0038  0.0133   0.2854  0.7753  -0.0223   0.0299     
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
summary(m_rep_age_interaction)
## 
## Multivariate Meta-Analysis Model (k = 79; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -158.5885   317.1770   327.1770   338.7644   328.0466   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.2139  0.4625     22     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 75) = 394.6953, p-val < .0001
## 
## Test of Moderators (coefficients 2:4):
## QM(df = 3) = 7.9638, p-val = 0.0468
## 
## Model Results:
## 
##                                  estimate      se     zval    pval    ci.lb 
## intrcpt                            0.4284  0.3537   1.2114  0.2257  -0.2647 
## mean_age                          -0.0000  0.0004  -0.0474  0.9622  -0.0007 
## n_repetitions_sentence             0.0336  0.0266   1.2628  0.2067  -0.0186 
## mean_age:n_repetitions_sentence   -0.0000  0.0000  -1.2990  0.1939  -0.0001 
##                                   ci.ub 
## intrcpt                          1.1216    
## mean_age                         0.0007    
## n_repetitions_sentence           0.0858    
## mean_age:n_repetitions_sentence  0.0000    
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

young, <36

m_rep_age_young <- rma.mv(d_calc ~ mean_age + n_repetitions_sentence , V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_young)
m_rep_age_interaction_young <- rma.mv(d_calc ~ mean_age*n_repetitions_sentence , V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_young)
summary(m_rep_age_young)
## 
## Multivariate Meta-Analysis Model (k = 57; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -114.6909   229.3817   237.3817   245.3376   238.1980   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.3048  0.5521     17     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 54) = 296.5847, p-val < .0001
## 
## Test of Moderators (coefficients 2:3):
## QM(df = 2) = 0.5337, p-val = 0.7658
## 
## Model Results:
## 
##                         estimate      se     zval    pval    ci.lb   ci.ub 
## intrcpt                   0.4734  0.4891   0.9681  0.3330  -0.4851  1.4320    
## mean_age                 -0.0002  0.0005  -0.3465  0.7290  -0.0012  0.0008    
## n_repetitions_sentence    0.0072  0.0149   0.4859  0.6271  -0.0220  0.0364    
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
summary(m_rep_age_interaction_young)
## 
## Multivariate Meta-Analysis Model (k = 57; method: REML)
## 
##    logLik   Deviance        AIC        BIC       AICc 
## -112.7315   225.4630   235.4630   245.3145   236.7396   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.3118  0.5584     17     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 53) = 293.9522, p-val < .0001
## 
## Test of Moderators (coefficients 2:4):
## QM(df = 3) = 3.5403, p-val = 0.3156
## 
## Model Results:
## 
##                                  estimate      se     zval    pval    ci.lb 
## intrcpt                           -1.3840  1.1765  -1.1764  0.2395  -3.6899 
## mean_age                           0.0024  0.0016   1.5307  0.1258  -0.0007 
## n_repetitions_sentence             0.1412  0.0786   1.7958  0.0725  -0.0129 
## mean_age:n_repetitions_sentence   -0.0002  0.0001  -1.7343  0.0829  -0.0004 
##                                   ci.ub 
## intrcpt                          0.9219    
## mean_age                         0.0056    
## n_repetitions_sentence           0.2952  . 
## mean_age:n_repetitions_sentence  0.0000  . 
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

old > 36

m_rep_age_old <- rma.mv(d_calc ~ mean_age + n_repetitions_sentence , V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_old)
m_rep_age_interaction_old <- rma.mv(d_calc ~ mean_age*n_repetitions_sentence , V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_old)
summary(m_rep_age_old)
## 
## Multivariate Meta-Analysis Model (k = 22; method: REML)
## 
##   logLik  Deviance       AIC       BIC      AICc 
## -41.2518   82.5035   90.5035   94.2813   93.3607   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.0000  0.0000      6     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 19) = 99.0894, p-val < .0001
## 
## Test of Moderators (coefficients 2:3):
## QM(df = 2) = 8.2020, p-val = 0.0166
## 
## Model Results:
## 
##                         estimate      se     zval    pval    ci.lb    ci.ub 
## intrcpt                   1.0411  0.4503   2.3119  0.0208   0.1585   1.9237   * 
## mean_age                 -0.0003  0.0003  -1.1152  0.2648  -0.0009   0.0003     
## n_repetitions_sentence   -0.0515  0.0181  -2.8419  0.0045  -0.0871  -0.0160  ** 
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
summary(m_rep_age_interaction_old)
## 
## Multivariate Meta-Analysis Model (k = 22; method: REML)
## 
##   logLik  Deviance       AIC       BIC      AICc 
## -40.8223   81.6445   91.6445   96.0964   96.6445   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.0000  0.0000      6     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 18) = 97.1759, p-val < .0001
## 
## Test of Moderators (coefficients 2:4):
## QM(df = 3) = 10.1155, p-val = 0.0176
## 
## Model Results:
## 
##                                  estimate      se     zval    pval    ci.lb 
## intrcpt                           -1.4947  1.8876  -0.7918  0.4285  -5.1944 
## mean_age                           0.0017  0.0015   1.1273  0.2596  -0.0013 
## n_repetitions_sentence             0.7744  0.5974   1.2964  0.1948  -0.3964 
## mean_age:n_repetitions_sentence   -0.0007  0.0005  -1.3833  0.1666  -0.0016 
##                                   ci.ub 
## intrcpt                          2.2050    
## mean_age                         0.0046    
## n_repetitions_sentence           1.9453    
## mean_age:n_repetitions_sentence  0.0003    
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

vocab

vocab (median)

ma_data_with_vocab <- ma_data %>% 
  mutate(vocab = case_when(!is.na(productive_vocab_median) ~ productive_vocab_median,
                           !is.na(productive_vocab_mean) ~ productive_vocab_mean,
                            TRUE ~ NA_real_),
         vocab_source = case_when(!is.na(productive_vocab_median) ~ "median",
                           !is.na(productive_vocab_mean) ~ "mean",
                            TRUE ~ NA_character_)) 

ma_data_with_vocab %>%
  ggplot(aes(x = productive_vocab_median, y = d_calc)) +
  geom_point() +
  geom_smooth(method = "lm") +
  theme_classic()

m_vocab <- rma.mv(d_calc ~ productive_vocab_median, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_with_vocab)
summary(m_vocab)
## 
## Multivariate Meta-Analysis Model (k = 36; method: REML)
## 
##   logLik  Deviance       AIC       BIC      AICc 
## -76.1570  152.3139  158.3139  162.8930  159.1139   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.2602  0.5101      6     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 34) = 160.5343, p-val < .0001
## 
## Test of Moderators (coefficient 2):
## QM(df = 1) = 6.7042, p-val = 0.0096
## 
## Model Results:
## 
##                          estimate      se     zval    pval    ci.lb    ci.ub 
## intrcpt                    0.8285  0.2557   3.2401  0.0012   0.3273   1.3297 
## productive_vocab_median   -0.0063  0.0024  -2.5892  0.0096  -0.0111  -0.0015 
##  
## intrcpt                  ** 
## productive_vocab_median  ** 
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

age comparison

ma_data_with_vocab_age <- ma_data_with_vocab %>% 
  filter(vocab_source == "median" &
        !is.na(vocab)) 

ma_data_with_vocab_age %>% 
  ggplot(aes(x = mean_age, y = d_calc)) +
  geom_point() +
  geom_smooth(method = "lm") +
  theme_classic()

m_age_with_vocab <- rma.mv(d_calc ~ mean_age, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_with_vocab_age)
summary(m_age_with_vocab)
## 
## Multivariate Meta-Analysis Model (k = 36; method: REML)
## 
##   logLik  Deviance       AIC       BIC      AICc 
## -76.1407  152.2813  158.2813  162.8604  159.0813   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.2767  0.5261      6     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 34) = 161.6858, p-val < .0001
## 
## Test of Moderators (coefficient 2):
## QM(df = 1) = 7.1130, p-val = 0.0077
## 
## Model Results:
## 
##           estimate      se     zval    pval    ci.lb    ci.ub 
## intrcpt     0.9727  0.2867   3.3922  0.0007   0.4107   1.5347  *** 
## mean_age   -0.0006  0.0002  -2.6670  0.0077  -0.0010  -0.0002   ** 
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

< 36, young

ma_data_with_vocab_age_young <- ma_data_with_vocab_age %>% 
  mutate(age_months = mean_age/30.44) %>% 
  filter(age_months < 36) 


ma_data_with_vocab_age_young %>% 
  ggplot(aes(x = productive_vocab_median, y = d_calc, size = n_1)) +
  geom_point() +
  geom_smooth(method = "lm") +
  ylab("Effect Size") +
  xlab("Age (days)") +
  ggtitle("Syntactical Bootstrapping effect size vs. vocab (months)") 

CATEGORICAL_VARS <- c("sentence_structure", "agent_argument_type", "patient_argument_type", "stimuli_actor",
                     "presentation_type", "character_identification",
                     "test_mass_or_distributed", "practice_phase", "test_method")


ma_data %>% filter(!is.na(mean_age)) %>% summarize(mean_age = mean(mean_age)) 
## # A tibble: 1 x 1
##   mean_age
##      <dbl>
## 1     907.
ma_data_ageExpl <- ma_data %>% mutate(AGE_group = (if_else(mean_age > 915.22, "OLD", "YOUNG")))
 

ggplot(ma_data_ageExpl, aes(x = sentence_structure, 
                    y = d_calc, 
                    group = presentation_type, 
                    color = presentation_type
                   )) +
  geom_line() +
  geom_point(position = "jitter") +
  geom_point(alpha = .4)  +
  ylab("Effect Size") +
  xlab("Sentence Structure") +
  ggtitle("ME effect size by Sentence Structure") 

ggplot(ma_data_ageExpl, aes(x = sentence_structure, 
                    y = d_calc, 
                    group = unique_id, 
                    color = unique_id
                   )) + 
  geom_line() +
  geom_point(position = "jitter") +
  geom_point(alpha = .4)  +
  ylab("Effect Size") +
  xlab("Sentence Structure") +
  ggtitle("ME effect size by Sentence Structure") 

ma_data %>% 
  group_by(agent_argument_type_clean) %>% count
## # A tibble: 4 x 2
## # Groups:   agent_argument_type_clean [4]
##   agent_argument_type_clean     n
##   <chr>                     <int>
## 1 noun                         21
## 2 noun_phrase                   9
## 3 pronoun                      23
## 4 varying_agent                26
ggplot(ma_data, aes(x = agent_argument_type_clean, 
                    y = d_calc, 
                    color = agent_argument_type_clean)) +
  geom_violin() +
  geom_point(alpha = .4)  +
  ylab("Effect Size") +
  xlab("Response Mode") +
  ggtitle("ME effect size by Response mode") +
  theme_classic() +
  theme(legend.position = "none")

cis_by_aa <- ma_data %>%
    group_by(agent_argument_type_clean) %>%
    summarize(mean = mean(d_calc),
            sd = sd(d_calc),
            n = n()) %>%
    mutate(ci_range_95 =  1.96 * (sd/sqrt(n)),
         ci_lower = mean - ci_range_95,
         ci_upper = mean + ci_range_95)

#pdf("plots/moderator_plot1.pdf", height = 6, width = 6)

ggplot(ma_data, aes(x = agent_argument_type_clean, 
                    y = d_calc, 
                    color = agent_argument_type_clean)) +
  geom_violin() +
  geom_point(alpha = .4)  +
  ylab("Effect Size") +
  xlab("Response Mode") +
  ggtitle("ME effect size by Response mode") +
  geom_pointrange(data = cis_by_aa, 
                  aes(x = agent_argument_type_clean, 
                      y = mean, ymin = ci_lower, 
                      ymax = ci_upper), 
                  color = "black") +
  geom_hline(aes(yintercept = 0), linetype = 2) +
  theme_classic(base_size = 16) +
  theme(legend.position = "none")

< 36, young, age comparison

ma_data_with_vocab_age_young %>% 
  ggplot(aes(x = age_months, y = d_calc, size = n_1)) +
  geom_point() +
  geom_smooth(method = "lm") +
  ylab("Effect Size") +
  xlab("Age (days)") +
  ggtitle("Syntactical Bootstrapping effect size vs. Age (months)") 

m_age_with_vocab_young <- rma.mv(d_calc ~ mean_age, V = d_var_calc,
                            random = ~ 1 | short_cite, data = ma_data_with_vocab_age_young)
summary(m_age_with_vocab)
## 
## Multivariate Meta-Analysis Model (k = 36; method: REML)
## 
##   logLik  Deviance       AIC       BIC      AICc 
## -76.1407  152.2813  158.2813  162.8604  159.0813   
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed      factor 
## sigma^2    0.2767  0.5261      6     no  short_cite 
## 
## Test for Residual Heterogeneity:
## QE(df = 34) = 161.6858, p-val < .0001
## 
## Test of Moderators (coefficient 2):
## QM(df = 1) = 7.1130, p-val = 0.0077
## 
## Model Results:
## 
##           estimate      se     zval    pval    ci.lb    ci.ub 
## intrcpt     0.9727  0.2867   3.3922  0.0007   0.4107   1.5347  *** 
## mean_age   -0.0006  0.0002  -2.6670  0.0077  -0.0010  -0.0002   ** 
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

correlationt est

cor.test(ma_data_with_vocab$mean_age,
         ma_data_with_vocab$productive_vocab_median)
## 
##  Pearson's product-moment correlation
## 
## data:  ma_data_with_vocab$mean_age and ma_data_with_vocab$productive_vocab_median
## t = 11.46, df = 34, p-value = 3.181e-13
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  0.7957418 0.9435289
## sample estimates:
##       cor 
## 0.8912725